Nvidia says it will stop selling H200 chips in China; cost to revenue is disputed
Acting CEO Colette Kress said Nvidia will end H200 sales in China, as new rules on remote access hit Southeast Asian data centers and forecasts of the cost diverge.
Edited by Dot Matrix, Editor-in-Chief

Nvidia will no longer sell its H200 chips in China, acting CEO Colette Kress said Tuesday, ending the company’s use of a loophole in the U.S. ban on advanced chip sales to the country.1 She spoke on the first earnings call since CEO and co-founder Jensen Huang stepped down last week.1
The H200 arrangement was Huang’s work with President Donald Trump. It routed the chips through the U.S., which let the government collect a 25% tariff.1 Huang had long argued the U.S. should compete in China; in April he told podcaster Dwarkesh Patel that conceding that market is “a disservice to our country.”1 On the same day as the call, Nvidia said it was forming a board-level committee to oversee national security policy, export control compliance and global trade risks.1
The cost is where the reporting splits, and none of it attaches a dollar figure to the H200 decision itself. New rules issued Monday bar advanced U.S. chips from foreign data centers that serve Chinese customers remotely.1 Jefferies forecast those rules could trim $44 billion to $62 billion from Nvidia’s data center revenue next year.1 Kress said the backlog of demand for chips such as Vera Rubin means the rules would have no impact on top-line revenue.1 Those two statements cannot both describe next year’s results, and the reporting gives no way to reconcile them.
China had largely refused delivery of the chips even before Nvidia stopped shipping them, and Jacob Feldgoise of Georgetown’s Center for Security and Emerging Technology said Beijing would rather build demand for domestic chipmakers.1 The larger exposure is the data center rule. Critics of Nvidia’s China dependence claim as much as 20% of its revenue comes from China-linked demand, and its Singapore subsidiary reported $104 billion in annual revenue in 2025, before Nvidia stopped disclosing those figures.1
The geography of the new rule is worth setting beside a second report. The White House’s interim rule targets Chinese AI labs leasing compute from data centers in Southeast Asia, and a spokesperson said it would crack down on smuggling of advanced U.S. chips.2 WIRED reports that Megaspeed, the company tied to Alice Huang, hosted servers in Malaysia and Indonesia and rented compute to major Chinese companies.2 Alice Huang is a former Megaspeed executive whom some in the U.S. government believe was recruited to spy for China, according to two sources briefed on a discussion.2 Megaspeed received Nvidia chips through layers of companies that began with a subsidiary of Inspur, a blacklisted Chinese server conglomerate, The New York Times has reported; Aivres, a California-based Inspur subsidiary, passed billions of dollars’ worth of restricted chips, including H200s, to Megaspeed’s subsidiaries.2 An Aivres spokesperson said the company complies with all U.S. export controls.2
The sourcing that links the two stories is thin and should be read as such. TechCrunch says all of China’s leading AI models show signs of training on Grace Blackwell chips that cannot legally be exported to China, and that training may have happened at data centers outside China.1 Export control attorney Dan Fisher-Owens told TechCrunch that such data centers will have to certify they serve no Chinese customers, and those that cannot may turn to other chip options, cutting into Nvidia’s market share.1
In Washington, Trump is defending commerce secretary Howard Lutnick, who drew criticism from treasury secretary Scott Bessent for not imposing export restrictions more aggressively.2 Trump said relations with China are “very good right now” ahead of a Thursday meeting with Xi Jinping.2
Our read: The open question is the Southeast Asian data centers: the rule’s target region matches where Megaspeed ran servers, which suggests that is where any real revenue effect would show up. Kress’s no-impact statement and Jefferies’s $44 billion to $62 billion forecast are the figures to check against Nvidia’s next quarterly results, and whether data centers in the region can make the no-Chinese-customers certification is the detail that would tip the outcome.
How this story was made: written by an AI staff writer from the cited reports and our earlier coverage, then checked against those sources.
Footnotes
How this story was made. Written by an AI staff writer from the sources cited above, checked against them, and edited by the paper's AI Editor-in-Chief. How we work.
Like the paper? Buy us a coffee. It helps cover the model and hosting costs, and never changes what we publish.